Sistemi Distribuiti Distribuiti I Corso di Laurea in Ingegneria

Transcript

Sistemi Distribuiti Distribuiti I Corso di Laurea in Ingegneria
Dipartimento di Sistemi e Informatica, University of Florence
Sistemi Distribuiti I
Corso di Laurea in Ingegneria
Prof. Paolo Nesi
Part 19 – Overview of social Network
Department of Systems and Informatics
University of Florence
Via S. Marta 3, 50139, Firenze, Italy
tel: +39-055-4796523, fax: +39-055-4796363
Lab: DISIT,
DISIT Sistemi Distribuiti e Tecnologie Internet
[email protected] [email protected]
http://www.dsi.unifi.it/~nesi
http://www.disit.dsi.unifi.it, http://www.axmedis.org
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
1
Overview of Social Network







Definition of Social Network
Terminology and Social Networks
Classification of Social Networks
User Generated Content, UGC
Measures of Social Networks
Recommendations and complexity
XMF:: to view inside a social network
XMF
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
2
Sistemi Distribuiti , Prof. Paolo Nesi
1
Dipartimento di Sistemi e Informatica, University of Florence
Introduction

With the users demand in collaborating and sharing
information some Social Networks have been created

Social Networks are (OECD, Organisation for Economic Co
Co-operation and Development
Development)) web portals that:
 Allow users to provide and share User Generated Content
 Allow users to valorize their creative effort, the content should be
originally produced by the users, take a picture, compose a set
of images, sync. images and audio, etc.
 Allow users to produce content by using solutions and non
professional techniques

Other solutions using UGC are Blogs, Wiki, Forum, etc.
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
3
Terminology

Social Network
 A paradigm of user interaction and behavior on the web

Social Media
 A Social network based on media

Social TV
 A TV based on Social Networking principles, with the support of
UGC, etc.

Social Network Analysis
 The discipline to analyze the social network in terms of user
clustering and relationships
relationships, metrics for SN assessment,
assessment etc
etc..
 It can be used to better understand motivation and rationales of
success and/or problems.
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
4
Sistemi Distribuiti , Prof. Paolo Nesi
2
Dipartimento di Sistemi e Informatica, University of Florence
Classification of Social Networks

Content Based Social Network:
 Collect content and show them to users according to their
preferences
 Content correlation
 Content recommendations
 Examples: YouTube, Last.fm, Flickr

User Based Social Network :
 User collection, user profiled
 Audio and video are used to better describe the user profile, in
some cases
cases, they are only visible to their friends
 User Recommendations, taking into account a large number of
user description aspects
 Examples: FaceBook,
FaceBook, Orkut,
Orkut, Friendster

MySpace is a mix of both categories.
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
5
Votes/ranks, Comments, preferred

Users may leave on Content and Users:
 Comments
 Ranks and Votes

Comments may be left as
 Text or content

User may mark the preferred content and users (friends)
 Preferred content are accessible with a direct list to
shortening the time for their play
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
6
Sistemi Distribuiti , Prof. Paolo Nesi
3
Dipartimento di Sistemi e Informatica, University of Florence
User Generated Content, UGC

Conditions that Facilitated the grown of UGC
 Reduced costs for equipments which allow the personal content
production: cameras, smart phones, etc.
 Reduced costs of connection, increment of broadband diffusion
 More Web Interactive capabilities: Ajax, JSP
 Creative Commons Licensing/formalisms, increment of
confidence

Pros and Facilitations
 Growing of WEB sites that host your content and provide some
tools to make them accessible on web for your friends
 Natural selection/emergence of better UGC items, increment of
visibility for some of UGC users…
 Annotation and reuse of UGC of others
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
7
User Generated Content

Cons and problems
 Restricted social penetration since only IT skilled and a certain
economical capability may access now
 Lack of formal Privacy control
Too many information are requested
Some people do not expose their true personal info
 IPR problems:
Violation of IPR of third party, free usage of UGC
Lose of control of your own UGC
Reuse and annotation of professional content
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
8
Sistemi Distribuiti , Prof. Paolo Nesi
4
Dipartimento di Sistemi e Informatica, University of Florence
User Generated Content

Cons and problems
 Lack of interoperability for users and content among different
social networks:
Initially performed to keep connected the users
Secondly a point of cons since users tend to pass from one
SN to another
 Content is not completely defined in terns of Metadata
 Competitions of UGC against professional content, producers are
against their support and diffusion
 Growing
g costs for the SN p
providers
Content volume in the hand of the SN organizers is growing
• Users would like to see older content still accessible
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
9
User Activities on Social Networks

Wikipedia (2006)
 68000: active users
 32 millions of lurkers
 While the 1000 more active users produced the 66% of
changes.

Similar numbers in other portals:
 90% lurkers
 9% occasional users
 1% acti
active
e users
sers
 90% is produced by the 1% of active users
 10% is generated by the 9% of users including the occasional
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
10
Sistemi Distribuiti , Prof. Paolo Nesi
5
Dipartimento di Sistemi e Informatica, University of Florence
Social Network Activities meaning

Since the 90% is managed by a small percentage of
active users:
produced with the same small p
part of
 Votes are also p
the community
 Comments are also produced with the same small
part of the community
 Pushers are frequently needed to create activities
and waves into the Social Networks, they create
fashions and interests among the lurkers, etc.. ..

Number of plays are produced by the whole
community
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
11
The centrality of User profile

User Profile Static information






Name, surname, Nationality
Genre, age, languages, etc..other personal info,..
School, work, family, etc.
photo, etc..
Economical data
User Profile Dynamic Information
 Explicit Preferences in terms of content, friends, votes, ranks,
recommendations, etc..
 Actions:
Actions pla
play, comments
comments, votes,
otes
 Frequency of access
 Etc.
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
12
Sistemi Distribuiti , Prof. Paolo Nesi
6
Dipartimento di Sistemi e Informatica, University of Florence
Relevance of Users

Number of Connections with other users
Direct connections,
 Second and third level connections,
 Etc.
E

Number of accesses to their





profile page (if any)
posted and/or preferred content
Comments
groups
U
Users’
’ Activities
A ti iti




Number of Posted content
Number of posted comments
Number of votes, etc.
Number of accesses
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
13
Stanford Social Web
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
14
Sistemi Distribuiti , Prof. Paolo Nesi
7
Dipartimento di Sistemi e Informatica, University of Florence
Issues on Communitie Graphs

Presence of a main Center of gravity
 Presence of dense groups

Presences of remotely located smaller Groups
 Self connections among these people
 Some of these smaller remote groups are linked with the
rest via 1 or more chains of single people
Depending on their activities, there is a risk of losing
those communities is evident

Number of Connections
 Distribution of connections
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
15
Shortest path from one person to another
MIT: 6.4 hops
Stanford: 9.2 hops
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
16
Sistemi Distribuiti , Prof. Paolo Nesi
8
Dipartimento di Sistemi e Informatica, University of Florence
Metriche per le Social Network


Social Network Analysis
Degree of Centrality per un Nodo:
 Numero di collegamenti incidenti sul Nodo

Eccentricity of Centrarlity per un Nodo:
 La dist. massima fra le distanze minime di tale nodo e ogni altro
nodo della rete

Closeness of Centrality per un certo Nodo:
 Reciproco della somma delle distanze tra il nodo e tutti gli altri
nodi

B t
Betweenness
Centrality
C t lit per un certo
t nodo:
d
 Quanta informazione passa per quel nodo. Somme delle quantita’
di informazione che passa fra tale nodo ed ogni altro nodo della
rete.
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
17
Recommendations

They are a means for the





Different Recommendations






Usage of content/object info to find/propose users
Usage of users info to find/propose content
Usage of users info to find/propose other users
Etc..
U  U: a user to another user on the basis of his profile
O  U: an object at a user on the basis of his profile
O  O: an object on the basis of a played object of a user
G  U: a group to a user
Etc…
Objects can be Advertising, Ads, Content, Events,
Groups, etc.….
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
18
Sistemi Distribuiti , Prof. Paolo Nesi
9
Dipartimento di Sistemi e Informatica, University of Florence
Different Recommendations
FOR YOU: Suggesting Users to another Users since they







have similar preferences
like/prefer what you like/prefer
are friends of your friends
are in one or more of the your groups
are new of the SN!
are the most linked, the most grouped, etc.
FOR THE SN: Suggesting Users to another Users since they

 are important
i
t t for
f the
th SN and
dd
do nott h
have tto lleft
ft alone,
l
the
th new entry
t
 are the only contact path for Connecting a remote group, if the path is left a
peripheral group will be completely disjoined with respect to the rest of the SN
 …
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
19
Complexity of Recommendation

Each day N new users reach the SN,
The SN has to suggest its possible friends immediately:





1 Million of users in the SN (number of users, U=10^6)
N*U distances to be estimated in real time/per day
Complexity is an O(NU)
Thus: 10^12 estimations of 10ms, thus 10^10s, 317 years !!!
Each day M new UGC items are posted on the SN,
The SN has to estimate the distance of that content with respect
to all the other items/objects and users:





1 Million of content in the SN (number of content,
content C=10^6)
M*C distances to be estimated in real time/per day
M*U distances to be estimated in real time/per day
Complexity is an O(MC+MU)
Thus: 10^12 estimations of 10ms, thus 10^10s, 317 years !!!
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
20
Sistemi Distribuiti , Prof. Paolo Nesi
10
Dipartimento di Sistemi e Informatica, University of Florence
XMF: CrossMediaFinder
ricerche
Contributi
Attivita’
Sociali
Carica contenuti
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009

Utenti e Servizi:





21
registrazione via email
email,, profilo utente, …
ricerche di altri utenti per stabilire relazioni sociali, …
upload di contenuti, User Generated Content
Content,, UGEsperiences,
UGEsperiences, …
conversioni automatiche dei loro contenuti per la distribuzione
multicanale, …
Aspetti Sociali, Social Network:
 commenti su contenuti, creazioni di discussioni sui contenuti, etc.
 gestione Contenuti Preferiti, visione dei contenuti caricati/preferiti
da/di amici,
amici …
 gestione dei propri Amici, Gruppi (ancora non attivo), …
 Produzione raccomandazioni per trovare altri amici
 Produzione raccomandazioni per trovare contenuti, … (ancora non
attivo), …
 ……
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
22
Sistemi Distribuiti , Prof. Paolo Nesi
11
Dipartimento di Sistemi e Informatica, University of Florence
Visualizzazione di Suggerimenti e dist
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
23
Factory and integration
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
24
Sistemi Distribuiti , Prof. Paolo Nesi
12
Dipartimento di Sistemi e Informatica, University of Florence
Architettura del XMF Social Network
Per la realizzazione è stato usato il Content Management System Drupal, integrato con l’applicazione realizzata in tecnologia J2EE XMediaFinder.
Drupal: Linguaggio PHP •Linguaggio PHP/Database Mysql
/
b
l
•Struttura modulare
•Gestione utenti: registrazione, permessi, profili
•Gestione contenuti: nodi.
•libreria Javascript JQuery
Applicazione XMediaFinder: applicazione che gestisce i contenuti Axmedis Fornisce pagine per:
Axmedis. Fornisce pagine per:
• Visualizzare le liste di oggetti più/meno visti, più/meno votati
•Ricercare (in modalità semplice e avanzata) i contenuti
•Visualizzare un contenuto
•Effettuare l’upload di un contenuto
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
25
Architettura del XMF Social Network
Schema di inclusione della pagine di XMediaFinder:
Inclusione Ajax
Tipo di nodo “Axmedis object”:
•Esiste un nodo “Axmedis obj” per ogni oggetto Axmedis
•Aggiunge funzionalità e informazioni di Drupal agli oggetti Axmedis:
commenti, votazioni, utente che ha effettuato l’inserimento.
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
26
Sistemi Distribuiti , Prof. Paolo Nesi
13
Dipartimento di Sistemi e Informatica, University of Florence
Moduli Drupal realizzati per il
XMF Social Network
•Modulo Axmedis: gestisce l’interazione con l’applicazione XMediaFinder
e i nodi di tipo “Axmedis object”
•Modulo User Information: gestisce gli utenti e la loro visualizzazione.
•Package Amedis User Information: i moduli in questo package
gestiscono i profili degli utenti, aggiungendovi le informazioni necessarie.
27
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
SN Comparison on Users
YouTube
Flickr
User profile, descriptors
Y
Y
Y
Y
Y
Friends
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Query on Users
FaceBook LikedIn MySpace
XMF
Y
Groups and Forums
Y
Y
Y
Y
Y
Y
Multilingual pages
Y
Y
Y
Y
Y
--
Invitations of users
Y
Y
Y
Y
Y
Y
Chats, on line, messages
Y
Y
Y
Y
Y
N
Recommendation UU
User Relevance, User,Obj,Group
N
N
Y
Y
Y
Y
Y(UO)
Y(OG)
Y(UG)
Y(UG)
Y(UG)
N
Y(G)
User Lists,
Lists gen rec.
rec of users
Y
N
Y
Y
Y
Taxonomy on Users
N
N
N
N
N
Y
Direct call, SMS, Email
Y
Y
Y
Y
Y
Y(SE)
Privacy support, Black List users
Y
N
Y
Y
Y
N
Events
N
N
Y
Y
Y
N
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
28
Sistemi Distribuiti , Prof. Paolo Nesi
14
Dipartimento di Sistemi e Informatica, University of Florence
SN Comparison on Content
YouTube
Flickr
Y(M)
Y(M)
Y(M)
N
N
Y(MC)
Audio, Video, Images, Doc
V
I, V
I, D, V
I, D
I, V
A,V,I,D
Moderated UGC
Y
N
N
(Y)
Query on content
Y
Y
N
N
Y
Y
Comments on Content
Y
Y
--
--
Y
Y
Ranking and voting
Y
N
--
--
Y
Y
General Recommendation O
Y
Y
Y
Y
Y
Y
Recommendation OU
Y
Y
--
--
Y
Y
Recommendation OO
Y
N
--
--
N
(Y)
Taxonomy for content/profile
N
N
N
N
N
Y
Play Lists of content
Y
N
N
N
N
N
RSS Feeds for content
Y
Y
Links with other SN
Y
Y
Y
Y
Multimedia, crossmedia UGC
Mobile Support
DRM/CAS Support
GeoTagging
FaceBook LikedIn MySpace
XMF
Y
N
Y
(Y)
Y
Y
Y
Y
Y
Y
Y(D)
N
N
N
N
Y (D)
Y
Y
N
N
N
N
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
29
Links For Further Research










Flickr: photo sharing community http://www.flickr.com/
Flickr:
YouTube: video sharing community http://www.youtube.com/
Myspace.. www.myspace.com
Myspace
Facebook.. www.facebook.com
Facebook
Friendster. www.friendster.com
Orkut.. www.orkut.com
Orkut
CrossMediaFinder,, XMF. http://xmf.axmedis.org/
CrossMediaFinder
Last.FM: social networking through music interests http://www.last.fm/
create your own social network http://www.ning.com/
MOODLE: open source ee-learning system http://moodle.org/
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2008-2009
30
Sistemi Distribuiti , Prof. Paolo Nesi
15